Footer

Your go-to destination for discovering trending open-source projects and uncovering the insights that matter

Legal

  • Terms and conditions
  • Privacy policy

More

  • Pricing
  • Advertise
  • onedump
  • gosnakego
  • Contact us

Social

  • X (Twitter)
  • GitHub
Trendshift owl
What they build speaks louder than words
© 2026 Trendshift, created by Julian Li with ❤️
SignalAdvertise
Trendshift
SignalAdvertise
Sign in
Trendshift owl

Discover trends that matter

SignalAdvertise
DailyWeeklyMonthlyYearly
Live mentionsTopics
RepositoriesDevelopers
InsightsStats

Bifrost
Bifrost
OfinTech Desk
OfinTech Desk
Busbar
Busbar
It's a plan
It's a plan
Omnigraph
Omnigraph

invergent-ai/surogate

Train and serve LLMs at extreme speed and massive throughput.

#LLM training
Visit GitHub
C++
888
16
4 contributors
Apache License 2.0
website
AchievementsView all →
FeaturedAdvertise to 180K+ monthly visitors
Bifrost

Fastest Enterprise AI Gateway for Routing, Governing, and Securing all your AI Traffic

OfinTech Desk

An open-source AI workspace for finance and accounting teams, in which every figure is traced to its source

Busbar

Control what AI can do before it acts

It's a plan

Open-source Linear alternative

An issue tracker where AI agents take the work you assign them

Omnigraph

Versioned graph database on object storage. Git-style workflows and multimodal retrieval

Momentum

How fast invergent-ai/surogate is gaining stars, and where it lands on Trendshift and GitHub Trending

Activity and trending

Stars, forks and merged pull requests each day or month, lined up with the repository's rank on Trendshift Trending

Sep 22 – Nov 19, 2026

Stars +75 in 59 days

0510
9

Forks +6 in 59 days

01
1

Merged PRs 138 in 59 days

0102030
25

Trendshift Trendingwhere the repository ranked each day it made the top 25

Not in the Trendshift top 25 in this period

Sep 22Sep 30Oct 8Oct 16Oct 24Nov 1Nov 9Nov 17
The left and right arrow keys step from day to day; Home and End go to the first and the last

Stars, forks, merged pull requests and Trendshift Trending rank, by day
DayStars gainedForks gainedMerged PRs
Tue, Sep 22, 2026+0+02
Wed, Sep 23, 2026+0+02
Thu, Sep 24, 2026+2+025
Fri, Sep 25, 2026+2+012
Sat, Sep 26, 2026+4+08
Sun, Sep 27, 2026+7+18
Mon, Sep 28, 2026+4+04
Tue, Sep 29, 2026+4+02
Wed, Sep 30, 2026+3+10
Thu, Oct 1, 2026+4+00
Fri, Oct 2, 2026+3+00
Sat, Oct 3, 2026+1+04
Sun, Oct 4, 2026+2+015
Mon, Oct 5, 2026+3+04
Tue, Oct 6, 2026+0+03
Wed, Oct 7, 2026+0+04
Thu, Oct 8, 2026+1+05
Fri, Oct 9, 2026+0+01
Sat, Oct 10, 2026+1+01
Sun, Oct 11, 2026+0+01
Mon, Oct 12, 2026+0+02
Tue, Oct 13, 2026+0+03
Wed, Oct 14, 2026+0+00
Thu, Oct 15, 2026+1+00
Fri, Oct 16, 2026+0+03
Sat, Oct 17, 2026+0+02
Sun, Oct 18, 2026+0+03
Mon, Oct 19, 2026+0+03
Tue, Oct 20, 2026+1+01
Wed, Oct 21, 2026+0+02
Thu, Oct 22, 2026+0+04
Fri, Oct 23, 2026+0+01
Sat, Oct 24, 2026+0+10
Sun, Oct 25, 2026+0+00
Mon, Oct 26, 2026+0+00
Tue, Oct 27, 2026+3+00
Wed, Oct 28, 2026+0+01
Thu, Oct 29, 2026+0+00
Fri, Oct 30, 2026+0+01
Sat, Oct 31, 2026+9+10
Sun, Nov 1, 2026+5+13
Mon, Nov 2, 2026+7+00
Tue, Nov 3, 2026+3+01
Wed, Nov 4, 2026+0+00
Thu, Nov 5, 2026+1+07
Fri, Nov 6, 2026+0+00
Sat, Nov 7, 2026+0+00
Sun, Nov 8, 2026+1+10
Mon, Nov 9, 2026+1+00
Tue, Nov 10, 2026+0+00
Wed, Nov 11, 2026+1+00
Thu, Nov 12, 2026+0+00
Fri, Nov 13, 2026+0+00
Sat, Nov 14, 2026+1+00
Sun, Nov 15, 2026+0+00
Mon, Nov 16, 2026+0+00
Tue, Nov 17, 2026+0+00
Wed, Nov 18, 2026+0+00
Thu, Nov 19, 2026+0+00
Total+75+6138
  • 1#1 on the list
  • 7Rank that day
  • Not trending

Mentions and achievements

Where people are talking about invergent-ai/surogate, and every badge it has earned
  1. Nov 192 posts
    • 18.8 s per agent turn vs 7.1 s. Same RTX 5090. The difference: 8 GiB of host RAM. 10 agent chats growing to 100k tokens, 4 at a time. When others need the GPU, a waiting agent's KV cache is dropped…

      Read full on X, posted by @flaviusburca

      @flaviusburca · x.com

    • Same GGUF model, same RTX 5090, 8 users at once: 4.05x apart. Qwen3.5-0.8B Q4_K_M, card at a 400 W limit, aggregate decode: Surogate 2,765 tok/s llama.cpp 683 tok/s Running a small local model for a…

      Read full on X, posted by @flaviusburca

      @flaviusburca · x.com

  2. Nov 182 posts
    • 53,900 vs 21,300 tokens/s. Same H100, same LoRA fine-tune. That's Surogate vs Unsloth on Qwen3-0.6B, both in BF16: 2.53x. The gap depends on model size. At Qwen3-8B, BF16 vs BF16, it's 10,200 vs…

      Read full on X, posted by @flaviusburca

      @flaviusburca · x.com

    • 19% apart on tokens/sec. 5.8x apart on time to first token. Same load test: 100 users, one RTX 5090 each, Qwen3.5-4B NVFP4, the same checkpoint in both engines. Surogate: 5,345 tok/s decode, 40 ms…

      Read full on X, posted by @flaviusburca

      @flaviusburca · x.com

  3. Nov 171 post2 posts
    • 839 ms vs 1,785 ms before the first word appears. Qwen3.8-27B NVFP4, one DGX Spark, 32 users at once, 512-token prompts. Median time to first token, Surogate vs vLLM: 2.13x. That wait is what your…

      Read full on X, posted by @flaviusburca

      @flaviusburca · x.com

    • 2.53x the training throughput of Unsloth: Qwen3-0.6B, BF16 on both, one H100 (53,900 vs 21,300 tok/s, our test). The engine is open source, so you can read the code and rerun it.

      Read full on X, posted by @surogate_ai

      @surogate_ai · x.com

  4. Nov 162 posts
    • If "sol" stays under 60% for a whole fine-tune, look for the bottleneck before you keep paying for GPU hours. sol is the speed-of-light number on every Surogate training step line: ideal step time on…

      Read full on X, posted by @flaviusburca

      @flaviusburca · x.com

    • About 16 s vs over 50 s just to read a cold 32k-token prompt. Qwen3.8-Flash-Next on one DGX Spark, prefill speed: - Surogate: 2,008 tok/s - llama.cpp: 629 tok/s That's 3.2x. Prefill is the model…

      Read full on X, posted by @flaviusburca

      @flaviusburca · x.com

  5. Nov 151 post
    • 2.4 s vs 7.2 s to the first token, on every warm agent turn. Gemma 4 26B-A4B on one DGX Spark, 1 agent at 100k-200k tokens of context, p50. Surogate vs vLLM 0.31. At 4 agents: 3.5 s vs 8.4 s. An…

      Read full on X, posted by @flaviusburca

      @flaviusburca · x.com

  6. Nov 141 post
    • An MoE load imbalance of 10 can be completely normal. Load imbalance = tokens sent to the busiest expert ÷ the average expert's tokens. 1.0 means perfectly even. But a trained router specializes, and…

      Read full on X, posted by @flaviusburca

      @flaviusburca · x.com

You might also like

Repositories close to invergent-ai/surogate
FeaturedAdvertise to 180K+ monthly visitors
It's a plan

Open-source Linear alternative

An issue tracker where AI agents take the work you assign them

Busbar

Control what AI can do before it acts

OfinTech Desk

An open-source AI workspace for finance and accounting teams, in which every figure is traced to its source

Bifrost

Fastest Enterprise AI Gateway for Routing, Governing, and Securing all your AI Traffic

Omnigraph

Versioned graph database on object storage. Git-style workflows and multimodal retrieval

microsoft/sarathi-serve

Python

A low-latency & high-throughput serving engine for LLMs

53969

naklecha/simple-llm

Python

~1300 line, minimal, extensible LLM inference engine built from scratch.

49338

noonghunna/club-3090

Python

Community recipes for serving LLMs on RTX 3090/4090/5090 CUDA gpus. Multi-engine (vLLM, sglang, llama.cpp and other custom engines) and model-agnostic. Currently shipping Qwen3.8-27B Qwen3.6 35B Gemma 4 26B Gemma 4 31B configs for 1×/2×/4x/8x cards.

2.5k185

iamwavecut/ninfer-all

C++

One line of NInfer for the Blackwell, Ada Lovelace, Ampere, Turing and Volta NVIDIA GPUs, consolidated from the forks that carry it: GGUF block formats, ternary Bonsai 2, device route profiles, MTP and DFlash2.

23451

lightseekorg/tokenspeed

Python

TokenSpeed is a speed-of-light LLM inference engine.

2.9k465

unslothai/unsloth-zoo

Python

Utils for Unsloth https://github.com/unslothai/unsloth

359338

vllm-project/vllm

Python

A high-throughput and memory-efficient inference and serving engine for LLMs

97.2k24.9k

harleyszhang/rapid_llm

Python

RapidLLM is an LLM inference framework with continuous batching, tensor and data parallelism, quantized weights, and with pluggable Triton/CUDA kernels.

19833

Lightning-AI/litgpt

Python

20+ high-performance LLMs with recipes to pretrain, finetune and deploy at scale.

13.8k1.7k